1

Mathematical modelling of pedestrian crashes: Review of pedestrian models

and parameter study of the influence of the sedan vehicle contour

Astrid Linder, Anthony Clark, Clay Douglas, Brian Fildes
Monash University Accident Research Centre (MUARC), Victoria, Australia

Jikuang Yang
Chalmers University, Sweden

Laurie Sparke
Holden Innovation, Victoria, Australia


INTRODUCTION
Road crashes result in a substantial number of pedestrian fatalities and injuries
worldwide. Statistics from 35 European countries have shown that pedestrian
fatalities represented on average 25% of road users killed throughout Europe
(ECMT, 2003). In Japan, pedestrian fatalities accounted for 28% of the road toll
(ECMT, 2003), while in Australia approximately 16% of road fatalities were
pedestrians (ATSB, 2003). Pedestrian fatalities as a proportion of road fatalities
were estimated at 13% in the USA to as high as 40-50% of the annual road toll in
India and Thailand (Mohan and Tiwari, 2000). The most common cause of such
fatalities is head injuries sustained by the pedestrian. Other severe injuries sustained
in such impacts are injuries to the chest, spine and abdomen and the lower
extremities (Anderson and McLean, 2001, Fildes et al., 2004). Computer simulations
for studying methods of reducing the loading to the pedestrian in a crash is a
powerful tool especially considering the great cost associated with mechanical
optimization of designs. Numerous pedestrian computer models are described in the
literature. These models vary in complexity, published validation and availability. For
the study of overall human kinematics in a crash, computer models based on rigid
bodies connected to each other by joints provide a powerful tool.

Pedestrian multibody models described in the literature can be divided into two
categories, models with and without frangible legs. In the category of frangible leg
models two widely used models are the Chalmers Pedestrian Model (Chalmers PM)
(Yang et al., 1997) and the TNO Pedestrian Model (TNO PM) (TNO, 2004). In the
category of non-frangible leg models there are a larger number of models, such as
the JARI Pedestrian Model (JARI PM) (Neale et al., 2001), the University of Adelaide
Pedestrian Model (Garrett, 1998) and the Honda Pedestrian Model (Honda PM)
(Okamoto et al., 2000). The UA PM is also referred to as the RARU i.e. Road
Accident Research Unit of Adelaide University, model in the literature in, for
example, Mizuno (2003). Figure 1 shows these models as 50
th
percentile male. All
the above mentioned models, except for the Honda PM, require a MADYMO working
environment. The Honda PM is run on PAM-CRASH. Validations of these pedestrian
models are performed against various cadaver tests. A comparison of the validation
of these models was not part of this review, as each validation focuses on a different
set of parameters thus limiting the ability to compare them.
2

Chalmers
Pedestrian Model
(Yang et al., 1997)
TNO Pedestrian
Model
(TNO, 2001)
JARI
Pedestrian Model
(Neale et al.,
2003)
University of Adelaide
Pedestrian Model (Garrett,
1998) (picture from Mizuno,
2003)
Honda Pedestrian
Model
(Okamoto et al.,
2000)
Figure 1 Multibody models of 50
th
percentile male pedestrians.

Pedestrian models with frangible leg
The Chalmers PM consists of 15 ellipsoids (Yang and Lvsund, 1997) and frangible
legs with a human-like knee. The human-like knee comprises a number of major
knee ligaments. The model can therefore be used to assess how aggressive the car
front is to pedestrians. The Chalmers PM has been validated against cadaver
experiments by Yang and Lvsund (1997) and Yang et al. (2000) and is not
commercially available. In addition to the 50
th
percentile male, this model has been
scaled to represent a 95
th
percentile male, a 5
th
percentile female and a 3, 6, 9 and
15-year-old child (Liu and Yang, 2002). The 50
th
percentile male pedestrian and the
9 and 6-year-old child models have been used in reconstructions of real-world
crashes by Yang (2003).
The TNO PM consists of 52 rigid bodies and includes six frangible leg joints in
each leg (TNO 2001). An overview of the validation of the TNO PM model against
cadaver experiments is found in TNO (2001). The TNO PM can be scaled to any
required body size (TNO 2001) and is commercially available from TNO. The TNO
PM is available in the MADYMO database and represents a 95
th
and 50
th
percentile
male, a 5
th
percentile female and a 6 and 3-year-old child. The 50
th
percentile male
model and the scaling of the TNO PM have been used in the reconstruction of real-
world crashes by van Rooij et al. (2003).

Pedestrian models without frangible leg
The most common type of multibody pedestrian model is that without frangible legs.
In this category there are the JARI PM, UA PM and the HONDA PM, with an
ongoing developmental process for most of these models. All these models
represent a 50
th
percentile male and none are commercially available.
The JARI PM comprises 27 segments whose dynamic response has been
compared to that of cadavers by Neale et al. (2003). The JARI PM has been used as
the basis for further improvements by IHRA (Neale et al., 2003) and resulted in the
development of the IHRA-PED Pedestrian Model (Mizuno, 2003).
The UA PM consists of 17 rigid segments, whose dynamic response has been
compared to that of cadavers in several studies, by among others Garrett (1998) and
Anderson and McLean (2001). The UA PM has been used in the reconstruction of
real-world pedestrian impacts by, for example, Garret (1998), Anderson and McLean
(2001), and Anderson et al. (2003) and as a result has been scaled to the
appropriate anthropometric data (segment dimensions, masses and motion of
inertia) for each pedestrian case.
3
The HONDA PM comprises 15 rigid bodies whose dynamic response has also
been compared to that of cadavers by Yoshida et al. (1998). The HONDA PM has
been scaled to represent, in addition to the 50
th
percentile male, a 95
th
percentile
male, a 5
th
percentile female and a 6-year-old child (Okamoto et al., 2000).

Previously performed optimization for a generic vehicle shape has shown that with
one vehicle parameter kept fixed, head and pelvic loads can be reduced (Niederer
and Schlumpf, 1984). Various vehicle shape parameters, such as the front of the
hood edge and the bumper lead, have been shown to affect in the loading to the
pedestrian in a crash (Higuchi and Akiyama, 1991) in such a way that the variation
of one parameter can be influenced by the other. Furthermore, the hood edge height
has been shown to influence the kinematics of the pedestrian (Ishikawa et al., 1991)
and the head responses to impact (Yang and Lvsund, 1997). Simulations by Yang
(2003) indicate that head impact conditions were affected by bumper lead and
height, hood leading edge height, hood length, hood angle and windscreen angle for
a child and an adult pedestrian model in impact speeds of 15 to 60 km/h. Corridors
representing sedans, light trucks and sports vehicles have recently been published
(Mizuno, 2003). Using mathematical modelling, the optimum vehicle contour for
minimizing the load to the pedestrian within such corridors could be investigated.


AIM
The aim of this study was to optimize the front contour of a sedan vehicle in order to
minimize the loads to the head of the pedestrian in a vehicle-pedestrian crash.


METHODS AND MATERIALS
As an initial step in the investigation of the influence of vehicle-specific features on
the loading to the pedestrian in a crash, the influence of the contour of a sedan
vehicle was examined using mathematical modelling. In this pilot study an
optimization of the contour of the vehicle was obtained for three different pedestrian
head output measurements.

Pedestrian Model
The pedestrian model used for this study was the Chalmers Pedestrian Model of a
50
th
percentile male pedestrian (Yang et al., 1997) and is run in MADYMO 5.4. This
model was made available for the simulations through a collaborative research
project aiming to improve pedestrian protection. The pedestrian model was
positioned in a walking stance, angled away from the vehicle at 15. The left upper
leg (vehicle strike side) was angled forward 6, with the right upper leg angled 5
backwards. The arms were positioned out of phase with the legs in human gait
cycles (Figure 2).


Figure 2 Pedestrian model (Yang et al., 1997) in the posture used in the optimization.
4
Design of the Mathematical Vehicle Model
The shape of the sedan model was optimization to fit within the sedan, light and
sports vehicle corridors developed by Mizuno (2003), which included 33 vehicles. As
a starting point for the simulations, a baseline vehicle profile that represented an
average large Australian passenger vehicle was designed. To represent this type of
vehicle, modifications were required to the hood length and windshield rake of the
average vehicle within the Mizuno (2003) corridor (Figure 3). In addition, the vehicle
was given a smooth aerodynamic shape. Furthermore, the bumper was divided into
an upper and lower section to represent the split bumper for radiator cooling found in
most vehicles. The vehicle geometry was then allowed to shift within pre-defined
limits of the corridors defined by Mizuno (2003). Furthermore, the upper and lower
bumper was allowed to move outside the IHRA corridor as there was very restricted
motion for these components allowed for within the corridor.

The curvature of the hood was constructed from a series of eleven planes using the
software Easi-Crash. Three cylinders were used to model the hood edge and
bumper systems. Contact interaction between the large leg ellipsoid and the small
hood edge plane required the use of a hood edge cylinder to allow appropriate
contact forces to be transferred, as previously observed by Ishikawa et al. (1993).
An additional plane was placed between the hood and the windshield to represent
the stiffer scuttle while a single flat plane was used to represent the vehicle
windshield. To prevent unrealistic multiple contact forces from occurring, the
evaluations command was applied where multiple contact interactions were
expected. Friction coefficients were applied to the pedestrian to vehicle contact (0.3)
and the pedestrian to ground contact (0.7), as well as damping.

Vehicle stiffness values were obtained from Yang et al. (2001). The hood edge
cylinder and corresponding planes were given a higher stiffness value than the hood
to represent the large box cross member located directly below the hood leading
edge (Yang et al., 1997). In addition to the stiffnesses in Yang et al. (2001) a
difference between upper and lower bumper stiffness was used since it was
considered that the lower bumper is typically less well supported, and hence a lower
stiffness value was applied (Figure 4). Furthermore, the higher hood edge stiffness
value was applied to the scuttle plane to represent a strike to the head on the stiff
scuttle, relative to that of a softer hood. In the simulations where the scuttle plane
was impacted, additional simulations were performed with a model of the hood that
continued up to the windshield so that the impact involved structures with the same
stiffness.



Figure 3 IHRA vehicle front shape corridor
(Mizuno et al., 2001) and the average large
Australian vehicle shape used as the baseline
vehicle shape.
Figure 4 Vehicle stiffness from Yang et al. (2001)
and the scuttle plane and upper and lower bumper
values used for the simulations.
IHRA Vehicle Profiles
Average Australian large passenger vehicle
Corridors
5

Vehicle components such as A-pillars, engine and strut towers, known to be
pedestrian unfriendly hard-points, were not modelled in this simulation for
simplification. Vehicle stiffness was not modified throughout the scope of this
analysis. In addition, no deceleration or nose-dive due to braking was included in the
vehicle dynamics.

The mathematical optimization program
A non-stochastic optimization program was designed to generate the optimum
vehicle contour for one pedestrian output parameter at a time. The optimization
program was written in MATLAB, as schematically described in Figure 5. The
program was designed to reduce the pre and post processing time required to
manually optimize vehicle structures. Designed as an automated logic optimization
program, the program optimizes geometry based on one selected pedestrian model
output for each optimization. The program was designed to minimize the output
parameter. One output parameter was optimized at a time due to that the complexity
of tracking multiple variables was beyond the scope of this pilot study.


Figure 5 Schematic structure of the mathematical optimization program

The optimization program evaluated a given geometric configuration based on the
increase or reduction of the pedestrian model output, after which the program made
decisions as to the next geometry modification. The program worked sequentially on
each vehicle component, such as lower bumper optimization then upper bumper
optimization. Each component was iterated a user specified number of times or until
the solution converged. In this way, the program cycled between 20-50 geometric
solutions before reaching an optimized solution for any given model output.

The optimization was carried out at the impact velocity of 40 km/h since at this
impact velocity the probability of fatal injuries for the pedestrian is significant (among
others Mizuno (2003), and Fildes et al. (2004)). In addition the baseline contour was
run at the impact velocity of 30 km/h to quantify the influence of impact velocity
compared with the influence of the contours found in the optimization. The
optimization program was run by minimizing three different head parameters; head
6
to vehicle resultant velocity ratio, head linear acceleration and head angular
acceleration. Within each optimization, peak outputs were recorded and presented
as a ratio to the baseline vehicle shape. All recorded peak values were checked for
consistency through graphs filtered at CFC 1000 and were only filtered at higher
rates (CFC 180) if erroneous spikes were observed in the time history file of the
output data.


RESULTS
The results of the simulations show the potential for designing the contour of a
vehicle in so as to reduce the loads to various body parts of the pedestrian in the
event of a pedestrian crash.

For the impact velocity of 40 km/h, the optimization of minimum linear and angular
head acceleration resulted in a reduced head velocity of 6% and 3% respectively
and a reduced HIC value by 81% and 53% respectively (Table 1) compared to that
of the baseline vehicle contour. These optimizations resulted in the largest increase
of pelvic loads and accelerations. When the contour minimizing head velocity was
created, head linear and angular acceleration as well as HIC increased. The largest
reduction in resultant head velocity was 13% compared to the baseline contour,
which was obtained when optimizing for minimum head resultant velocity.

Table 1 The pedestrian output parameters from the baseline vehicle and the ratio of the pedestrian
parameters for each optimization as a ratio to the baseline vehicle shape for the Optimizations at 40
km/h impact velocity.


40 km/h
Baseline
Vehicle
Optimizing for Minimum Head
to Vehicle Resultant Velocity
Ratio
Optimizing for
Minimum Head Linear
Acceleration
Optimizing for
Minimum Head Angular
Acceleration
HIC 16 3115 1.46 0.19 0.47
Head Velocity Resultant Ratio 1.24 0.87 0.94 0.97
Head Linear Acc. (g) 215 1.41 0.49 0.70
Head
Head Angular Acc. (rad/s
2
) 19103 1.37 0.56 0.78
Pelvis Pelvis Impact Force Value (kN) 5.65 0.96
*
3.70 3.28
*
Filtered with CFC 180 instead of CFC 1000



Optimised shape
Baseline vehicle
Vehicle optimised for
minimum head angular
acceleration
Optimised shape
Baseline vehicle
Vehicle optimised for
minimum head angular
acceleration

Optimised shape
Baseline vehicle
Vehicle optimised for
minimum head impact
velocity ratio
Optimised shape
Baseline vehicle
Vehicle optimised for
minimum head impact
velocity ratio
Optimised shape
Baseline vehicle
Vehicle optimised for
minimum head linear
acceleration
Optimised shape
Baseline vehicle
Vehicle optimised for
minimum head linear
acceleration
Figure 6 The vehicle contours generated for the impact velocity of 40 km/h in the optimizations
compared to the baseline vehicle shape that was used as a starting point for the simulations.

In the animation of the optimizations (Figure 7) it was found that in the case of the
optimized vehicle shape on head velocity, the head struck the scuttle plane and
therefore this case included head interaction with a stiffer structure than the other
cases. This particular case was rerun with a hood continuing up toward the
7
windshield to facilitate a direct comparison between all contours generated in this
study (Table 2).

Figure 7 The animations of the optimized cases and the base line vehicle at the impact velocity of 40
km/h.
Time Baseline Head res. velocity Head linear acc. Head angular acc.
70
ms



110
ms



130
ms



150
ms




Table 2 The pedestrian output parameters from the base line case and the ratio of these
parameters for the optimization for the impact velocity of 40 km/h with the hood continuing to the
windshield as a ratio to the baseline vehicle shape.


40 km/h
Baseline Vehicle Optimizing for Minimum Head to Vehicle Resultant Velocity Ratio
HIC 16 3115 0.83
Head Velocity Resultant Ratio 1.24 0.87
Head Linear Acc. (g) 215 0.94
Head
Head Angular Acc. (rad/s
2
) 19103 0.87
Pelvis Pelvis Impact Force Value (kN) 5.65 0.96
*

*
Filtered with CFC 180 instead of CFC 1000



The influence of the impact velocity for the baseline vehicle is shown in Table 3. In
this table the pedestrian output parameters from the 30 km/h impact are divided by
those from the 40 km/h impact.

Table 3 The pedestrian output parameters from the baseline vehicle and the ratio of these
parameters for each optimization for the impact velocity of 30 km/h divided by the results from the
40 km/h baseline simulation.
Pedestrian output parameters HIC
16
Head resultant
velocity
Head
linear acc.
Head angular
acc.
Pelvis
Force
Ratio between baseline
30 km/h and 40 km/h case
0.36 0.93 0.69 0.68 0.23

8
DISCUSSION

The results of the mathematical optimization of the vehicles contour indicate that
loading to the head of the pedestrian can be reduced with alteration to the design
within the corridors of an average sized vehicle. Both angular and linear acceleration
of the head and HIC were considerable reduced for a number of the contours
generated, indicating a potential for reducing these parameters by the contour of the
vehicle. Some designs identified in this study reduced the loads to the head and
distributed them to the pelvis.

The largest influence on resultant head velocity from these simulations was a
reduction of 13% from the baseline vehicle contour while the other optimizations had
minor influence on the resultant head velocity. These results indicated that within the
IHRA corridor for light trucks, sport and sedan vehicles (Mizuno, 2003), major
reductions in resultant head velocity at 40 km/h for a 50
th
percentile male pedestrian
might be limited. The contour generated when minimizing the resultant head velocity
resulted in the pedestrian model moving closest to the vehicle during the impact.
The two other optimizations caused the pedestrian to become airborne before
striking the head on the vehicle and therefore considerable loads to the neck and
spine could be expected. In these cases there was little or no chest impact (Figure
7).

The reduction of impact velocity has been shown to have a significant effect on
injury severity for children (Lui and Yang, 2002). This study shows similar results for
a 50
th
percentile male when the contour of the vehicle was optimized within the IHRA
corridor. Compared to the optimizations of vehicle contours at 40 km/h (Table 1 and
2), the head and pelvic output parameters were reduced for the baseline vehicle
contour when the impact velocity was reduced from 40 to 30 km/h (Table 3).

Due to the non-stochastic approach of the optimization program, a minimum for the
chosen output parameter was found. This approach will not necessarily identify the
absolute minimum, but a local minimum. Therefore, additional minima than those
identified from the optimization could exist. This is exemplified by the results in the
optimization for minimum head angular acceleration at 40 km/h, where a local
minimum for head angular acceleration was found. However, an even lower
minimum for head angular acceleration was identified in the optimization for the
minimum head linear acceleration, as shown in Table 1. For some of the impacts,
oscillations occurred in the output signals. This required increased filtering to CFC
180 to remove anomalies generated by these oscillations (Table 1) and highlighted
the importance of closely examining the time history events from mathematical
simulations.

A conflicting design phenomenon was observed for competing injury parameters. In
the 40 km/h study, minimum head velocity was generated by an extended lower
bumper, whereas minimum head acceleration requires the lower bumper to be as
the baseline vehicle. The minimum head velocity was generated by a lowering and
shifting forward of the hood, as opposed to minimum head angular and linear
acceleration, which was obtained by a retraction and rising of the hood (Figure 6).

9
The optimization program was originally developed to allow changes to the shape of
the hood. This resulted in the hood forming corrugations to accept the models
thorax, pelvis and head ellipses to successfully distribute the load as shown in
Figure 8. It was necessary to control this, and resulted in the hood geometry
confined to a continues line without the corrugations.

Figure 8 The corrugations of the hood that occurred in the originally developed program
that allowed individual changes in the orientation of the planes of the hood.

The influence of vehicle front design and the size of pedestrians have been studied
by Yang (2003). In this study it was shown that of the two different sedan models
used in the simulations, one of the models reduced the head impact velocity relative
to the vehicle for all the tested pedestrian sizes with one exception. The influence of
pedestrian size should be taken into consideration in order to make
recommendations about optimum vehicle design and will be aimed at in future
optimizations. Further optimizations should also include various vehicle front
stiffnesses, a larger range of impact velocities and lower extremity measurements.
Such Optimizations would be needed in order to make recommendations on the
vehicle design with the largest potential for reducing critical loads to the pedestrian
in a vehicle-pedestrian crash.


CONCLUSIONS

Reducing the head resultant velocity by optimizing the contour of a sedan vehicle
within the IHRA profile, at a 40 km/h impact velocity, resulted in a maximum
reduction of 13 % compared to the baseline vehicle contour. The largest reduction in
HIC was 81 %, the head linear acceleration was 51% and the angular acceleration
of the head was 44 %. The reduction of pedestrian head output parameter by
changing the contour of the vehicle increased in some of the cases the pelvic load.


REFERENCES

Anderson, R.W.G., Streeter, L., Ponte, F., Sommariva, M., and McLean, A.J. (2003). Vehicle
Design and Operation for Pedestrian Protection - Accident Simulations and
Reconstructions. Vehicle Safety Standards Report. Department of Transport and Regional
Services (DOTARS), Canberra, Australia, Report No VSSR 1, ISBN 0-642-50236-6.
Anderson R.W.G and McLean A.J. (2001) Vehicle design and speed and pedestrian injury:
Australias involvement in the International Harmonised Research Activities pedestrian
expert group, Road Safety Research, Policing and Education Conference Proceedings,
Melbourne, Australia.
ATSB (2003). Website: www.atsb.gov.au 17/09/03.
ECMT (2003) Statistical report on road accidents, ISBN 92-821-0299-8.
10
Fildes, B., Gabler, C.H., Otte, D., Linder, A., Sparke, L. (2004) Pedestrian Impact Priorities
Using Real-World Crash Data and Harm, Proc. of the International IRCOBI Conference on
the Biomechanics of Impacts. Graz, Austria.
Garrett, M. (1998). Head Impact Modelling Using MADYMO Simulations of Documented
Pedestrian Accidents. Proc. of the Conference on Pedestrian Safety. Melbourne, Australia.
Higuchi K. and Akiyama A. (1991) The Effect of the Vehicle Structures Characteristics on
Pedestrian Behaviour, 13
th
International Technical Conference on Enhanced Safety of
Vehicles, , Paris, France, Paper no 91-S3-O-10
Ichikawa H. Yamazaki K. Ono K. and Sasaki A. (1991) Current Situation of Pedestrian
Accidents and Research into Pedestrian Protection in Japan, 13
th
International Technical
Conference on Enhanced Safety of Vehicles (ESV), Paris, France, Paper no 91-S3-O-05.
Liu, X.J., and Yang, J.K. (2002). Development of Child Pedestrian Mathematical Modeland
Evaluation with Accident Reconstruction. Traffic Injury and Prevention, Vol. 3. pp. 321-
329.
Mizuno, Y. (2003). Summary of IHRA Pedestrian Safety WG activities (2003) Proposed
Test Methods to Evaluate Pedestrian Protection Afforded by Passenger Cars. 18
th

International Technical Conference on the Enhanced Safety of Vehicles (ESV), Nagoya,
Japan, Paper no 580.
Mohan D. and Tiwari G. (2000) Road safety in less motorized countries relevance of
international vehicle and highway standards, International Conference on Vehicle Safety,
IMechE, 7-9 June, London, Paper No C567/008/2000, pp 155-166.
Neale, M.S., Hardy, B.J., and Lawerence G.J.L. (2003). Development and review of the
IHRA (JARI) and TNO pedestrian models. Proc. of the 18
th
International ESV Conference,
Nagoya, Japan. Paper no 449.
Niederer P. F. and Schlumpf M. R (1984) Influence of Vehicle Front Geometry on Impacted
Pedestrian Kinematics, Stapp, 841663, pp 135-147
Okamoto, Y., Sugimoto, T., Enomoto, K. and Kikuchi, J. (2000). Pedestrian head impact
conditions depending on the vehicle front shape and its construction Full model
simulation, Proc. of the International IRCOBI Conference on the Biomechanics of Impacts.
Montpellier, France. pp. 281-291.
TNO (2001). Human Models Manual Version 6.0. TNO Automotive, Delft, The Netherlands.
van Rooij, L., Bhalla, K., Meissner, M., Ivarsson, J., Crandall, J., Longhitano, D., Takahashi,
Y., Dokko, Y., and Kikuchi, Y. (2003). Pedestrian crash reconstruction using multi-body
modelling with geometrically detailed, validated vehicle models and advanced pedestrian
injury criteria. Proc. of the 17
th
International ESV Conference, Nagoya, Japan. Paper 468.
Yang J (2003) Effect of Vehicle Front Design Parameters on Pedestrian Head-Brain Injury
protection, 18
th
International Technical Conference on Enhanced Safety of Vehicles (ESV),
Nagoya, Japan, Paper no 289.
Yang, J., Liu, X., Lovsund, P., Tingvall, C., Lie, A., Larsson, P., and Lekander., T. (2001).
Speed Limit in City Area and Improvement of Vehicle Front Design for Pedestrian Impact
Protection- a Computer Simulation Study. Proc. Of The International Technical
Conference on the Enhanced Safety of Vehicles (ESV 17). Amsterdam, The Netherlands.
Paper 232.
Yang, J.K., Lvsund, P. Cavallero C., Bonnoit J. (2000). A Human-Body 3D Mathematical
Model for Simulation of Car-Pedestrian Impacts. International Journal of Crash prevention
and Injury Control. Vol.2(2), pp 131-149.
Yang, J.K., and Lvsund, P. (1997). Development and Validation of a Human-Body
Mathematical Model for Simulation of Car-Pedestrian Impacts. Proc. of the International
IRCOBI Conference on the Biomechanics of Impacts. Hannover, Germany. pp. 133-149.
Yoshida, S, Matsuhashi, T. Matsuoka Y (1998). Simulation of Car-Pedestrian Accident for
Evaluate Car Structure. Proc. of the 16
th
International ESV Conference, Windsor, Canada,
Paper no 98-S10-W-18, pp 2344-2348.